Array-level MXene electrode flexible long-term-plasticity synaptic transistors

Abstract

Artificial nervous devices are expected to emulate the information processing methods of the human brain to achieve high-speed low-power computing. In these systems, synaptic transistors mimic the brain's input and storage mechanisms in response to external stimuli. Achieving a high-integration flexible synaptic transistor array with long-term plasticity and mass-production potential remains difficult due to structural and material limitations. In this work, novel array-level MXene-electrode flexible synaptic transistors have been realized with a long-term synaptic plasticity of over 1000 s without requiring a floating gate structure, electrolyte dielectrics, or ferroelectric materials. The Ti₃C₂T_x MXene electrodes provide protons from surface functional groups to fill the defects of the dielectric layer to achieve long-term synaptic plasticity. Furthermore, the method of patterning MXene electrodes enables both a high level of integration and array-level patterning, which had been highly challenging, especially on flexible substrates. Additionally, the matched work functions of MXene and semiconducting carbon nanotubes enable Ohmic contact for the metal–semiconductor junction. Furthermore, an MXene electrode flexible synaptic transistor (MEFST) array is made possible for next-generation 3D perception systems to achieve artificial neuromorphic computing and temporal fusion information storage based on long-term plasticity, leading to a high accuracy of 93.8% for recognising handwritten digits. Long-term plasticity allows rapid identification upon the reappearance of the same target without needing re-iteration. The MEFST array holds immense application potential for neuromorphic electronic skin, intelligent wearable electronics and edge computing.